What are world models — and are they Europe’s winning AI bet?

The AI debate has been dominated by large language models used to power the likes of ChatGPT, Claude and Gemini. But could the next wave of AI look very different?

Yes, some European AI watchers say. Attention is shifting to so-called “world models” — systems designed to build internal representations of how the world works, rather than simply predicting the next word in a sentence. 

A number of prominent researchers, including former Meta chief AI scientist Yann LeCun, have argued this approach could overcome some of the limitations of today’s LLMs — and have launched startups to prove it.

On this week’s episode of the Sifted Podcast, host Freya Pratty is joined by senior reporters Daphné Leprince-Ringuet and Anne Sraders to unpack the hype around world models. What exactly are they? What real-world applications might they unlock? And with researchers like LeCun choosing to base new ventures in Europe, could this be an area where the continent builds a competitive edge in the global AI race?

Sign up to our AI and deeptech newsletter here: https://sifted.eu/newsletters